Ordering only by the session's created date failed to realise we could have more than one entry for the same stable id. Ordering now by the score as a way of getting rid of this ambiguity. Issue was picked up by Rhoda running the code on an older MySQL version which consistently exhibited different behaviour to later MySQL versions

Switched to Test::More and done_testing(). This should stop tests from failing due to poor test count specification.

Test schema now matches live. Any future patching should be done using the ensembl-test/scripts/patch_test_databases.pl script.

use Test::More rather than old Test library. Avoids breaking tests when we change the numbers of tests run

[ENSCORESW-302]. TranslationAdaptor did not deal with multispecies databases properly when querying for IDs.

Added a no_cache override feature plus testcase. Useful in times of memory crisis coupled with annoyingly slow lookups of small data.

[ENSCORESW-271]. Fixed the test being too chatty during cleanup by setting the RUNTESTS_HARNESS_NORESTORE before firing off a thread

[ENSCORESW-264] Can now handle odd sequence region names involving . - _ now. The supercontig names seemed to be a good source of these

Making sure multiple includes of the registry does not result in major issues with error detection (earlier work suggested a possible issue with this)

We can now set a reconnection timeout outside of the normal realms of MySQL disconnect/reconnect. This means we have something closer to a pooled connection which stays active for as long as people need it & dies off at the required rate

Test cases now pass, we have a capture warns asserting for known bad scenarios and use the Test::Builder code to see if we want to do to screen warning for test cases. Not sure how un # lines in TAP are dealt with; we may want to substitute STDOUT/STDERR for something which can delegate to Test::Builder for note()

Trying to improve the current test cases. Accidentally caused some issues from rerunning the code multiple times as well

[ENSCORESW-224]. Fix for multi-line SQ lines caused by large sequences being pushed through SeqDumper. Was never an issue before as our slices were never big enough. Applied the same fix to Genbank but it should be less of an issue due to it's formatting meaning we need larger slices to cause the issue.

Tests null cases and basic functioning of sorting algorithm.

Adding the scope guard which is a way we can cause code to be executed even during a die. Very useful for non-global resource cleanup operations (finally blocks)

ENSCORESW-152. Committing the low cardinality and low row count caching framework into the core API. Implement the correct method from BaseAdaptor to have your values cached in local memory

Adding an iterator for subseqs to the Slice object. Means a lot of our code which deals with iteration of slices can delegate to this for its logic.

Additional patch to bring in version into the unique key for xref. Also apply changes to gene and transcript

Using the "best" HTTP client available i.e. avoid making people install LWP if they are on a late enough version of Perl. HTTP::Tiny is very nice